HPC - SGE Error

[Stephen D Lengyel]

Trinity Users,

I'm having an issue with HPC computing on SGE cluster. Trinity phase 1 completes, but entering phase 2 i get the following error:

"""--------------------------------------------------------------------------------
------------ Trinity Phase 2: Assembling Clusters of Reads ---------------------
--------------------------------------------------------------------------------

$VAR1 = {
          'grid' => 'SGE',
          'cmd' => 'qsub -cwd',
          'max_nodes' => '6',
          'cmds_per_node' => '1'
        };
Use of uninitialized value $job_id_text in concatenation (.) or string at /home/slengyel/Programs/Trinity/PerlLib/HTC/SGE_handler.pm line 43.

SGE:
Use of uninitialized value $job_id_text in concatenation (.) or string at /home/slengyel/Programs/Trinity/PerlLib/HTC/SGE_handler.pm line 49.
FARMIT failed to accept job: qsub -cwd /home/slengyel/B7H6/Sample1/trinity/farmit.J12376.itsusmaclun07.12376.1432914850/cmds/J12376.S0.sh 2>&1 """
 (ret -1)

I did notice that there wer conflicting "HTC" and "HPC" in the perl modules when running the SGE test in the PerlLIB directory. The test does succeed and the command that should fail, does.

Any idea what's going on here? I don't code in perl so I'm not sure why $job_id_text is giving an error.

All help is appreciated.

[Brian Haas]

Hi Stephen,

The job:

 qsub -cwd /home/slengyel/B7H6/Sample1/trinity/farmit.J12376.itsusmaclun07.12376.1432914850/cmds/J12376.S0.sh

wasn't able to submit, apparently.  

Are jobs running on SGE able to submit other jobs to SGE?  This is sometimes not allowed, as IT administrators sometimes turn off the functionality that would allow running jobs to submit other running jobs.

best,

~brian

--
You received this message because you are subscribed to the Google Groups "trinityrnaseq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trinityrnaseq-u...@googlegroups.com.
To post to this group, send email to trinityrn...@googlegroups.com.
Visit this group at http://groups.google.com/group/trinityrnaseq-users.
For more options, visit https://groups.google.com/d/optout.

--

--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

[Brian Haas]

ok - good luck!  

On Mon, Jun 1, 2015 at 10:08 AM, Stephen D Lengyel <tue8...@temple.edu> wrote:

Hi Brian,

It possible that is the case. The trininty SGE test worked without error. I"ll check the system admin to verify this.

Thanks,

Stephen

--

Stephen D. Lengyel, P.S.M.
PSM Liason, BGSS
Kulathinal Lab, 213A
Department of Biology
Temple University
stephen...@temple.edu

[Stephen D Lengyel]

HI Brian,

I contacted the admin, and he said that it should work as long as I have password-less ssh setup, which I was able to do. But I'm still getting the same error. The jobs aren't submitting for Inchworm after Phase I. Could there be an issue with the farmit pm?

-Stephen

[Brian Haas]

Hi Stephen,

To further test this, I'd suggest taking the original test that you were able to run successfully and see if it works if you qsub the test script.    That is, it appears to currently work from your login shell, but I'm curious if it will work if you submit it to SGE.  

Another thing you can do to test this is to simply make a shell script that has a 'qsub' command in it, and see if it works when you qsub that script.     If it doesn't work, then your IT administrator should help you figure out what's going on.  If it does work fine, then we can revisit Farmit, etc.

best,

~b

[Stephen D Lengyel]

I did a testing submitting a script to submit 3 other test scripts. Each test script contained a quick echo statment. The result was errors for all 3 qsubs within the original script "qsub: command not found." I have my .bashrc pointing to the location of the command. When i log into the node that the original test script was submitted, I can qsub.

[Brian Haas]

When you do your initial qsub, be sure to have '-V' as a parameter, as this should retain your environment upon submission, and the node running the script to do the other qsubs should be able to find the qsub command.

Let's see if that helps.

best,

~b

[Stephen D Lengyel]

Well that did. I've been so used to using PBS and I'm not quite adjusted to SGE.

Thanks, Brian!

-Stephen

[Brian Haas]

Great!  :)